Network Role Mining and Analysis

109 Pages·2017·2.079 MB·English

by  Derek Doran (auth.)

SPRINGER BRIEFS IN COMPLEXITY
Derek Doran
Network Role
Mining and
Analysis
123
SpringerBriefs in Complexity
Editorial Board for Springer Complexity
Henry Abarbanel, La Jolla, USA
Dan Braha, Dartmouth, USA
Péter Érdi, Kalamazoo, USA
Karl Friston, London, UK
Hermann Haken, Stuttgart, Germany
Viktor Jirsa, Marseille, France
Janusz Kacprzyk, Warsaw, Poland
Kunihiko Kaneko, Tokyo, Japan
Scott Kelso, Boca Raton, USA
Markus Kirkilionis, Coventry, UK
Jürgen Kurths, Potsdam, Germany
Andrzej Nowak, Warsaw, Poland
Hassan Qudrat-Ullah, Toronto, Canada
Linda Reichl, Austin, USA
Peter Schuster, Vienna, Austria
Frank Schweitzer, Zürich, Switzerland
Didier Sornette, Zürich, Switzerland
Stefan Thurner, Vienna, Austria
Springer Complexity
Springer Complexity is an interdisciplinary program publishing the best research
and academic-level teaching on both fundamental and applied aspects of complex
systems—cutting across all traditional disciplines of the natural and life sciences,
engineering, economics, medicine, neuroscience, social and computer science.
Complex Systems are systems that comprise many interacting parts with the
ability to generate a new quality of macroscopic collective behavior the mani-
festations of which are the spontaneous formation of distinctive temporal, spatial
or functional structures. Models of such systems can be successfully mapped onto
quite diverse “real-life” situations like the climate, the coherent emission of light
from lasers, chemical reaction-diffusion systems, biological cellular networks, the
dynamicsofstockmarketsandoftheinternet,earthquakestatisticsandprediction,
freeway traffic, the human brain, or the formation of opinions in socialsystems, to
name just some of the popular applications.
Although their scope and methodologies overlap somewhat, one can distinguish
the following main concepts and tools: self-organization, nonlinear dynamics,
synergetics, turbulence, dynamical systems, catastrophes, instabilities, stochastic
processes, chaos, graphs and networks, cellular automata, adaptive systems,
genetic algorithms and computational intelligence.
The three major book publication platforms of the Springer Complexity pro-
gram are the monograph series “Understanding Complex Systems” focusing on
the various applications of complexity, the “Springer Series in Synergetics”,
which is devoted to the quantitative theoretical and methodological foundations,
and the “SpringerBriefs in Complexity” which are concise and topical working
reports, case-studies, surveys, essays and lecture notes of relevance to the field. In
addition to the books in these two core series, the program also incorporates
individual titles ranging from textbooks to major reference works.
More information about this series at http://www.springer.com/series/8907
Derek Doran
Network Role Mining
and Analysis
123
Derek Doran
Department ofComputer Science
andEngineering
Wright State University
Dayton,OH
USA
ISSN 2191-5326 ISSN 2191-5334 (electronic)
SpringerBriefs inComplexity
ISBN978-3-319-53885-3 ISBN978-3-319-53886-0 (eBook)
DOI 10.1007/978-3-319-53886-0
LibraryofCongressControlNumber:2017932113
©TheAuthor(s)2017
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpart
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
orinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar
methodologynowknownorhereafterdeveloped.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom
therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authorsortheeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinor
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations.
Printedonacid-freepaper
ThisSpringerimprintispublishedbySpringerNature
TheregisteredcompanyisSpringerInternationalPublishingAG
Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland
Preface
Theadventoflarge-scaleonlinesocialnetworks,webservicesandsystemsthatare
inherently social, and our unprecedented ability to capture “big” meta-data about
entitiesincountlesssystemshaveledtoasurgeofrecentdevelopmentstodiscover
the roles of entities in networks by sociologists, mathematicians, statisticians, and
computer scientists. This monograph presents an overview of network role mining
and analysis techniques. It organizes methods into five classes, representing
methods of varying computational sophistication and interpretability. Techniques
within each class are presented with an eye towards their actual implementation in
computer algorithms and systems. The monograph also discusses the broad char-
acteristics of methods in each class, enabling high-level comparisons that guide
practitioners toward choosing the appropriate technique for a given analysis task.
Themonographtargetsresearchersandpractitionerswithlittletonobackgroundin
network role mining, those with experience in a limited number of methods, or
thosethatfeeltheyhavejustanarrowperspectiveofthiskindofnetworkanalytics.
The structure and components of almost any complex system can be examined
fromtheperspectiveoftherolesitscomponentsplay.Thismonographwillserveas
aguidepostforcomputationalsocialscientists,computerscientists,andstatisticians
toaidintheirunderstandingofthesemethods,tofindtherightkindofroleanalysis
tobeusedintheirproblemofinterest,andtoinspirenewresearchdirectionsinthis
important field.
Dayton, OH, USA Derek Doran
December 2016
v
Acknowledgements
I would like to thank my collaborators and students who engaged me in conver-
sations leading to the development of this monograph. Thank you to Kyle Brown,
JaceRobinson,andJibrilIkharofortheirhelpfuleditsandcommentsthatimproved
thereadabilityandcontentofthisbook.IespeciallywanttoacknowledgeKylewho
assisted me in various aspects of the blockmodeling chapters and engaged me in
manyconversationsconcerningthetopicsofthisbookbeforeitsconception.Thank
youalsotoChristopherCoughlin,whoseinterestingconversationsduringthe2015
IEEE/ACM International Conference at Social Network Analysis and Mining in
Paris, France provided the initial spark for this brief. Finally thanks to Zandra
Zweber, Dan Ford, and Andre Keuck for volunteering their time as professional
meteorological and extraordinary individuals consultants.
vii
Contents
1 Network Role Mining and Analysis: An Overview.. .... ..... .... 1
1.1 Introduction .... .... ..... .... .... .... .... .... ..... .... 1
1.2 Defining Roles .. .... ..... .... .... .... .... .... ..... .... 2
1.2.1 Networks. .... ..... .... .... .... .... .... ..... .... 3
1.2.2 Positions in Networks.... .... .... .... .... ..... .... 5
1.3 Mining Roles ... .... ..... .... .... .... .... .... ..... .... 6
1.3.1 Relationship to Graph Partitioning and Community
Detection. .... ..... .... .... .... .... .... ..... .... 10
1.4 Purpose and Outline of This Monograph ... .... .... ..... .... 11
References.. .... .... .... ..... .... .... .... .... .... ..... .... 11
2 Implied Role Mining. .... ..... .... .... .... .... .... ..... .... 15
2.1 Introduction .... .... ..... .... .... .... .... .... ..... .... 15
2.2 The Implied Role Mining Process .... .... .... .... ..... .... 17
2.3 Illustrations with Usenet.... .... .... .... .... .... ..... .... 18
2.3.1 Golder et al.’s Taxonomy. .... .... .... .... ..... .... 19
2.3.2 Nolker et al.’s Hierarchy.. .... .... .... .... ..... .... 22
2.4 Analysis of Implied Role Mining. .... .... .... .... ..... .... 25
2.4.1 Qualitative Nature... .... .... .... .... .... ..... .... 25
2.4.2 Compatibility . ..... .... .... .... .... .... ..... .... 26
2.4.3 Simplicity and Interpretability.. .... .... .... ..... .... 27
2.5 Conclusion. .... .... ..... .... .... .... .... .... ..... .... 29
References.. .... .... .... ..... .... .... .... .... .... ..... .... 30
3 Equivalence-Based Role Mining. .... .... .... .... .... ..... .... 31
3.1 Introduction .... .... ..... .... .... .... .... .... ..... .... 31
3.2 Structural Equivalence ..... .... .... .... .... .... ..... .... 32
3.2.1 Finding Structural Equivalences .... .... .... ..... .... 32
ix
x Contents
3.3 Automorphic Equivalence... .... .... .... .... .... ..... .... 34
3.3.1 Finding Automorphic Equivalences.. .... .... ..... .... 35
3.3.2 Quantifying Automorphic Similarity. .... .... ..... .... 37
3.4 Regular Equivalence.. ..... .... .... .... .... .... ..... .... 40
3.4.1 Finding Regular Equivalences.. .... .... .... ..... .... 40
3.4.2 Quantifying Regular Similarity. .... .... .... ..... .... 45
3.5 Conclusion. .... .... ..... .... .... .... .... .... ..... .... 46
References.. .... .... .... ..... .... .... .... .... .... ..... .... 46
4 Deterministic Blockmodeling ... .... .... .... .... .... ..... .... 49
4.1 Introduction .... .... ..... .... .... .... .... .... ..... .... 49
4.2 The Blockmodeling Framework .. .... .... .... .... ..... .... 52
4.2.1 Similarity Measures.. .... .... .... .... .... ..... .... 52
4.2.2 Blocktypes ... ..... .... .... .... .... .... ..... .... 56
4.3 Goodness of Fit . .... ..... .... .... .... .... .... ..... .... 58
4.3.1 A Goodness-of-Fit Measure for Positional Analysis .. .... 59
4.3.2 A Goodness-of-Fit Measure for Network Compression.... 60
4.4 Conclusion. .... .... ..... .... .... .... .... .... ..... .... 60
References.. .... .... .... ..... .... .... .... .... .... ..... .... 61
5 Stochastic Blockmodeling . ..... .... .... .... .... .... ..... .... 63
5.1 Introduction .... .... ..... .... .... .... .... .... ..... .... 63
5.2 SBM Specification ... ..... .... .... .... .... .... ..... .... 64
5.3 The Infinite Relational Model.... .... .... .... .... ..... .... 65
5.3.1 Parameter Inference for the IRM.... .... .... ..... .... 67
5.3.2 Summary. .... ..... .... .... .... .... .... ..... .... 70
5.4 The Dynamic Stochastic Blockmodel.. .... .... .... ..... .... 70
5.4.1 DSBM Network Generation ... .... .... .... ..... .... 71
5.4.2 Parameter Inference for the DSBM.. .... .... ..... .... 73
5.5 Conclusion. .... .... ..... .... .... .... .... .... ..... .... 75
References.. .... .... .... ..... .... .... .... .... .... ..... .... 75
6 Advanced Computational Methods .. .... .... .... .... ..... .... 77
6.1 Factor Graphs: The Social Roles and Statuses Factor Graph
Model. .... .... .... ..... .... .... .... .... .... ..... .... 78
6.1.1 Social Features ..... .... .... .... .... .... ..... .... 79
6.1.2 A Factor Graph Model ... .... .... .... .... ..... .... 80
6.2 Multi-view Learning: Dual-View Uncertainty Regularization. .... 82
6.2.1 Graph Co-regularization .. .... .... .... .... ..... .... 83
6.2.2 Uncertainty Regularization and Objective Function... .... 84
6.3 Bayesian Modeling: Co-discovery of Roles in Communities . .... 85
6.4 Matrix Factorization: RolX.. .... .... .... .... .... ..... .... 87
Contents xi
6.5 Iterative Quadratic Programming: Synergistic Co-discovery
of Communities and Roles.. .... .... .... .... .... ..... .... 89
6.5.1 Initializing Communities (InitCom).. .... .... ..... .... 90
6.5.2 Initializing Roles (InitRole).... .... .... .... ..... .... 90
6.5.3 Updating Communities... .... .... .... .... ..... .... 91
6.5.4 Updating Roles..... .... .... .... .... .... ..... .... 92
6.6 Conclusion. .... .... ..... .... .... .... .... .... ..... .... 92
References.. .... .... .... ..... .... .... .... .... .... ..... .... 93
7 Concluding Remarks. .... ..... .... .... .... .... .... ..... .... 95
7.1 Emerging Trends in Role Mining. .... .... .... .... ..... .... 96
7.2 Tension Between Rigor and Interpretability . .... .... ..... .... 98
References.. .... .... .... ..... .... .... .... .... .... ..... .... 100